Curating Flipboard content: two varieties of personalization

Curating my streams on Flipboard

The killer app

A few weeks ago, Flipboard (often referenced to as the iPad killer app”) was finally launched for the iPhone. And what the developers did to transport the flowing way of navigating your streams and news sources to the much smaller screen of the iPhone is more than convincing. Flipboard has very quickly become my favorite app for reading Twitter, Facebook or Instagram.

But there is one thing that really shows the future of curating news: With Flipboard everybody can generate his own stream of news-sources and make it available for other users. Actually, I believe Jon Russell, is wrong. He complained earlier this day, that

“the news is somewhat US/Western-centric for me. While it is true to say that, particularly when talking tech (though I am referring all news genres), the lion’s share of media is US based. But, if I want personalised news, and the chance to discover content, should there be more emphasis on varied sources?”

Two flavors of personalization

There are two different varieties of personalization. The “old” variety that surfaced with the first dynamic web pages based on CGI scripts, allowed the user to select from a number of options defined by the editor (or webmaster). This variety is based on the notion of more or less homogeneous audiences or large groups of people. In this world you could chose between European, North-American or Asian news sources the same way you could chose between Tech, Culture or Politics channels. But you could not create, edit or share these streams.

The “new” variety of personalization allows to do exactly that. You can not only chose your streams but create, edit and share your streams. In Flipboard the most convenient way to do this is via Twitter lists. You only have to create a Twitter list of people or sources (= bots) that reflect your information needs the best, add this list to your Flipboard and off you go with your personalized news ecosystem in the “new” variety. You can also share your news list by making the list public. This way you can not only curate content, but also curate interesting people.

Writing for Flipboard

Another option is to curate your own Flipboard channel via a Twitter account, Facebook account or a generic RSS source such as a blog. I believe that there will be many applications resembling Flipboard in the future, so that sooner or later it will make sense for media companies to have a “Flipboard curator” that knows how to write optimized for this kind of aggregators.

Almost 5000 tweets per minute during Obama’s inauguration

It is no big surprise that Obama’s inauguration generated a lot of buzz on Twitter. The hashtag #inaug09 quickly became one of the most frequently typed Twitter tag. But exactly how often did people tweet about Obama’s inauguration? We have quite detailed data about this day on Facebook:

Since this morning more than 1.5 million status updates have been posted through the feed (there were 200,000 b 8:30 am PST). During the broadcast an average of 4,000 status updates were written every minute, and 8,500 were written every minute during Obama’s speech.

Twitter only published a chart on the corporate blog – a chart without numbers:
Inauguration day on Twitter
They wrote:

We saw 5x normal tweets-per-second and about 4x tweets-per-minute as this chart illustrates.

But how much is 5x normal or 4x normal? Mashable wrote: “Tweets containing “Obama” hit 35,000 per hour during his speech.” This would be 583 Obama tweets per minute. Not too much.

My method: Fortunately every tweet has its own id which is simply increased by 1 with every new tweet. So to calculate the average Twitter speed during Obama’s inauguration, I looked for the first Obama tweet that had been posted at 12:00 and the first tweet at 1:00. The difference between those two is the number of tweets published during this hour.

The result: During 12pm and 1am, there had been 4746 tweets per minute (TPM).

TwitterFriends: Overlapping networks

One of the new features of TwitterFriends is a visualization of your incoming and outgoing networks. Remember: Your incoming network includes all Twitter users that replied or referred to you on Twitter the last 30 days at least twice. And the outgoing network – your relevant net – includes all Twitter users you referred or replied at least twice. You can see this new diagram on your stats page: Just replace kevinrose with your Twitter username. No password required.

Kevin Rose’s diagram (I chose him because he’s the Twitter user having the largest incoming network) looks like this:

You can see three things on this diagram:

  • Popularity: The network of people addressing him is much larger than the network of people he’s addressing. It’s almost eight times larger (see number in the left box of the statistics page). This measure goes in the same direction as the Friend-to-Follower-Ratio. If a lot of people are sending you replies and you are only able to reply to a small fraction of them, you’re popular. So, the different sizes of the two networks are a measure for popularity.
  • Overlap: The overlap is not too large. In fact it’s only 28.6% (see number in the left box of the statistics page). There are many people trying to talk to Kevin Rose on Twitter that are not receiving replies. But on the other hands there are Twitter users, Kevin Rose talks to that do not answer his replies regularly or their accounts are protected – TwitterFriends does not analyze protected data, even if you log in with your Twitter credentials at the top of the page.
  • Friends: If you want to know more about the people in the overlapping area, take a look at the full user cloud either incoming or outgoing.

    @kevinrose's Outgoing network

    Those names with a bidirectional arrow are contacts that are talking to Kevin Rose on a more or less regular basis and receiving Twitter messages by Kevin Rose as well. Of course, these users are not always what we’d are calling “friends” in real life, but they are regular conversation partners.

Now you can take a look at your own TwitterFriends statistics and compare your Venn diagram with the above. Are your networks more overlapping? Do you have a larger incoming than outgoing conversational network as well? If you experience a bug or have a great idea about how to improve TwitterFriends, send me a Twitter message to @furukama or just submit your input to the TwitterFriends support forum.

How efficient is your Twitter network?

Valdis Krebs just wrote a great blog post about his strategy of maximizing his Twitter network efficiency. We all know, the more people you are following on Twitter, the more difficult it gets to keep in touch with all of them. Most often you are reducing your network size by selectively reading about and replying to the people you really care about (the relevant net) or the people you are talking to you (reciprocity).

But if you are looking at your Twitter network as a informational network (and not so much as a relationship network), it is more important how your contacts are related. If you are using your network to keep informed about many different topics, it pays to build a heterogenous network of many people from wholly different areas of expertise. Valdis put it this way:

The trick is to find the people that reach many social circles and follow them. Of course, we need to find more than the minimum of people to follow — you want some redundancy in your network so that there are multiple paths to places of interest for you.

In Social Network Analysis (SNA), there are some ways to put this notion into quantitative metrics. Ron Burt described a measure of redundancy in his greatly acclaimed work on Structural Holes. Stephen Borgatti continued this line of thought and developed a set of metrics to describe redundancy, density and network efficiency. This last measure interested me because it provides a way to measure the effective size of your network – how it would look like without redundant nodes. I’ve written a short function for TwitterFriends to display this measure on the network tab:

Twitter Network Efficiency for @furukama

Twitter Network Efficiency for @furukama

After looking at some other Twitter users’ networks, @valdiskrebs’ network efficiency of 95.01% still ranks among the highest values. So, his strategy of building a diverse network that allows for information from different sources to reach him, seems to be quite successful.

To find out your Twitter Network Efficiency, visit this site: and replace the username with your Twitter username. Maybe the value will seem high at the beginning, but this can be because not all your friends’ connections are in the TwitterFriends database yet. Click on the names below the graphic to load them.

Visualizing Twitter networks

Now, I’ve finally rolled out the network visualization mode for TwitterFriends. It does not show the entire network of Twitter contacts (followees and followers) because it would simply be too large and confusing. Also, it would be not very meaningful, since contacts are easier added than removed. Therefore, only the “relevant network” of contacts, a person responded or addressed at least twice with the ‘@’ syntax will be visualized. This is the hidden network of people, a person is giving his best attention publicly on Twitter. After entering a Twitter user name, the ego network will be drawn. Here the one for my Twitter account @furukama:

@furukama's TwitterFriends network

@furukama's TwitterFriends network

The size of the nodes of the network corresponds to the number of Tweets, I have written to my Twitter contacts. Thus they represent the intensity of communication between the central or ego node and the relevant contacts in the network. As this visualization can become very crowded for users with many contacts, you can surf through this network: By clicking on a node, the network centers around it, allowing you a deeper look into the connections between the nodes in your relevant network:

Looking at @saschalobo's network in @furukama's network

Looking at @saschalobo's network in @furukama's network

Below the network graphic, there is a link for toggling between the simple network representation above and full FOAF visualization (FOAF for “friend of a friend“). The second visualization also includes the contacts of my contacts, which are not in my network – my “friends’ friends”. This increases the number of nodes once again, but you can click on nodes to navigate the network. Here’s the full network for Robert Scoble:

Robert Scoble's FOAF network on Twitter

@scobleizer's FOAF network

And here’s mine:

@furukama's FOAF network

@furukama's FOAF network

The network visualization requires no additional installed software such as Java or Flash. I’ve used the great JavaScript visualization library JIT by Nicolas Garcia Belmonte. The whole thing should run on Firefox, Safari and Chrome and with reduced speed in Internet Explorer. Visit to try it out with your own Twitter network.

Note: For some users the network may initially appear quite empty because the network data of their contacts are not yet cached. In this case, there will be a list of links below the network graph. By clicking on them you can collect the missing data. Users with private profiles or who have not used @replies can not be included. You do not need to enter your Twitter credentials to visualize your network.

Comments and bug reports via Twitter to @furukama or in the comments. Thank you!

Measuring relevant networks

See our friends,
see the sights,
feel alright

Relevant net

I’ve already mentioned the paper by Bernardo A. Huberman, Daniel M. Romero and Fang Wu about Twitter’s social networks (pdf). The point that interested me most was the distinction between contacts and friends. Especially for people with a liberal following policy (i.e. that follow back once someone follows them on Twitter), their network of contacts is not very meaningful. But what is a meaningful network on Twitter? Huberman, Romero and Wu propose that there is a hidden network of friends that you just have to extract from your overall contact network. This hidden network is transaction-based: it’s the people you have replied to using the @syntax more than once. Because it is based on the choices you made, I’d prefer to call it the “relevant net” because it works similar to the “relevant set” known from marketing. This network consists of the few people from your total network that you find meaningful enough to actively address them.

This is my relevant net (calculated with “TwitterFriends”) consisting of all my Twitter contacts that I addressed with the @syntax more than once in my last 1,000 twitter messages (click to enlarge):


@furukama's relevant net on Twitter

@furukama's relevant net on Twitter

Inner circle

My relevant net consists of 92 nodes.  In my last 1,000 tweets, I addressed only 92 of my 1,400 contacts (6.6%) more than twice. But the replies are not equally distributed. There are many people I only addressed twice and there are some people I talked to very often. The above visualization shows the different layers of my relevant net. In the middle is the “inner circle” of my Twitter friends: a small number of people I addressed very often. The colors are important: blue are people I met in real life, orange are people I have not (yet) met and gray are bots or organizations. My inner circle is blue. This means that the people I talk to most often on Twitter are people I also know and talk to in real life.

Network analysis

The next step is taking a look at my Twitter friends and see how they are connected. Again, I did not use the formal notion of Twitter contacts, but the more specific notion of Twitter friends, i.e. people who talked to each other more than twice. To produce this network visualization, I removed myself, because all of the nodes in my relevant net are connected to me (that’s how they are defined) and so this does not add any information to the graph. Here are the conversations between members of my relevant net (click to enlarge):


Network of my Twitter friends

Network of @furukama's Twitter friends

Although this conversational network is not as dense as the network resulting from the following/followers network, it is not, as you would expect from what Huberman, Romero and Wu are writing, exactly a sparse network. Its density is 6,06%. This means, that of all 8,372 potential directed connections, 507 are realized. 8 nodes have received no replies at all. Without them, density would be 7,45%. Often, for directed valued graphs like this one (= graphs differentiating between incoming and outgoing ties) density is calculated as the average of the line weights (= number of replies between two persons). In this case, the density is 39. This means: the average number of replies from one of my Twitter friends to another is 39.

The distribution of the incoming @replies in my network looks like a power distribution, which means that this network should be a scale-free network. Few people in my network receive a large number of replies while a large number receives only few replies.

Distribution of incoming @replies

Distribution of incoming @replies

What does this all mean?

  • Social Media Measurement definitely should take a closer look at social network analysis. If you want to find out who are the influencers in a network like Twitter, looking at the numbers of followers could be misleading. @having has a large number of followers and is one of the nodes in my relevant net that gets the most replies. Is it an influencer? No, it isn’t. It doesn’t talk back.
  • Twitter should find a method allowing us to keep closer in touch with our relevant net. This small subset of all my contacts is often what’s interesting me the most. I’d love to have a tool that keeps me updated on them. But I don’t want to select my relevant net manually as it is possible with Twitbin-Groups. Above I’ve described a way, this can be done automatically.
  • This is a good example for the importance of relevance instead of total reach (e.g. number of followers, visitors, clicks) in social media. The actual exchance of messages, the dialogue that’s happening on Twitter is taking place in such relevant nets. Phenomena like these deserve much more attention by social media marketeers.

If you want to know your relevant net, you can calculate your Twitter friends with this application.

Networks that matter on Twitter: the @-Crowd

Finally I found a research paper that analyzes Twitter from a network analytical perspective. The main message of the paper “Social networks that matter: Twitter under the microscope” by Huberman, Romero and Wu (see also Jeremiah Owyang’s summary) is: the number of people you follow on Twitter is not the whole truth. It’s more interesting who you are talking to whether you are following them or not. It not a connection-based network but a performance-based network. When taking a look at your Twitter network, there are three different kinds of networks:

  • The Network: simply the network of your followers / followings. Those are the people whose updates you might be reading and who might be receiving your updates. This is the reach of your Twitter stream.
  • The FOAF-Network: the network of your followers’ / followings’ networks. Those are the people you could potentially reach via retweeting messages. This is the extended reach of your Twitter stream.
  • The @-Crowd (or the “hidden network”): the people you are talking / replying to. This is the most interesting measure, because it’s the people you are explicitely addressing. The authors of the paper call contacts you have addressed (or replied to) at least twice “friends”.
So, let’s take a closer look at the people I am talking to on Twitter (= my @-Crowd) with a small tool I just wrote. It only shows Twitter users I sent at least 2 replies – i.e. “friends” in the terminology of Huberman, Romero and Wu.
You can see from this analysis of my latest 400 tweets that I’m using the @-syntax more often than the average Twitter user (25,4%). A larger part of my twitter messages actually is a dialogue with a small number of people. In my last 1,000 tweets, for example, I did only reply to 6% of my total Twitter contacts more than once.
If you take a look at different users (just replace the user=username with your Twitter username), you see that there are different patterns of Twitter usage. Some talk with very many different people, some use the @-syntax only in a few replies and some are using Twitter as a dialogue tool.

Towards a global database of social media marketing case studies

Which companies are using social media like blogs, social communities, internet forums or video/picture sharing for their marketing efforts? The first lists of case studies are being published right now on various blogs:

  • Peter Kim has a long list with over 300 brands’ social media efforts.
  • Then there are many case studies if you take a look at this category in Jeremiah Owyang’s blog.
  • Geoff Livingston has a selection of cases on his blog.
  • I collected German cases of social media marketing on my German language blog.

And now I’ve learned that Peter Kim and his community will publish the cases in “A Wiki of Social Media Marketing Examples“. Great idea! I will enter my cases when I have received my account for editing the wiki.

Anyway, what is still missing are numbers. There is a very lively debate on social media measurement (measuring influence, engagement, ROI etc.) going on – but there are no numbers, not even estimates in these lists at all. In my opinion, this lack of numerical evidence is what’s holding German companies back from spending more money for social media marketing. The BBS – big budget shift – will have to wait until we can offer the marketing executives a convincing story about which numbers for reach, engagement, performance they can expect in social media marketing. They know their numerical environment very well when it comes to TV or newspaper ads but this is not the case with social media. They have to be able to tell whether 5.000 installed widgets on Facebook are a success story or failure.

An important step will be international standards for measuring the different facets of social media. To encourage debate about definitions, standards and best practices, the German Social Media Association will host a “Social Media Measurement Summit” on 27 January 2009 in Munich, Germany. I’m looking forward to discussing these matters with you all in Munich!

What is Social Media? A not so critical review of concepts and definitions

In 1952 Alfred Kroeber and Clyde Kluckhohn published their seminal work “Culture: A Critical Review of Concepts and Definitions”. They provided a systematic overview of 164 definitions of this anthropological key concept. I did not find as much definitions of social media and furthermore the differences between definitions social media are far smaller than the differences between different concepts of culture.

What is Social Media?

This is a quick overview of the first 141623 definitions of social media by the following people:

  • Ben Parr: “… the use of electronic and Internet tools for the purpose of sharing and discussing information and experiences with other human beings in more efficient ways”
  • Anthony Mayfield: “… a group of new kinds of online media, which share most or all of the following characteristics: 1) Participation … 2) Openness … 3) Conversation … 4) Community … 5) Connectedness”
  • Clay Shirky: “… stuff that gets spammed”
  • Robert Scoble: “… Internet media that has the ability to interact with it in some way”
  • Brian Solis: “… put the power of media into the hands of the people, which transformed content consumers into content producers” or “Social Media is the democratization of information, transforming people from content readers into publishers. It is the shift from a broadcast mechanism to a many-to-many model, rooted in conversations between authors, people, and peers.” (see below)
  • Stowe Boyd: “1) Social Media Is Not A Broadcast Medium … 2) Social Media Is Many-To-Many … 3) Social Media Is Open … 4) Social Media Is Disruptive”
  • Chris Heuer: “… redefining how we relate to each other as humans and how we as humans relate to the organizations that serve us”
  • Joseph Thornley: “… online communications in which individuals shift fluidly and flexibly between the role of audience and author”
  • David Meerman Scott: “… online media with a participatory or interactive component”
  • Ike Pigott: “… strange brew of Technology + People + Organization + Freedom”
  • Deirdre Breakenridge (PR 2.0, FT Press 2008, xviii): “… anything that uses the Internet to facilitate conversations between people”
  • Robert Berkman (The Art of Strategic Listening, Paramount, 10): “Blogs, wikis, digital videos or any other kind of textual or multimedia forms of media and typically generated by ordinary consumers”
  • Beth Kanter: “… a way of using the Internet to instantly collaborate, share information, or have a conversation ideas or causes we care about”
  • Daniel Nations: “… a social instrument of communication”
  • Eric Karjaluoto: “… media that users can easily participate in and contribute to”
  • Louis Gray: “… a loose term that largely relies on user generated content”
  • Marta Kagan: “… people having conversations online”
  • Sarah Worsham: “… it’s about the shared meaning you create with your customers as you interact with them and they with each other”
  • Mark Dykeman: “… the means for any person to: publish digital creative content; provide and obtain real-time feedback via on-line discussions, commentary, and evaluations; and incorporate changes or corrections to the original content”
  • Ashwini Dhagamwar and Sandeep Arora: “… allows people to participate using Media (text, audio, video, pictures) instantly. Social Media is the story about people participation on a scale never seen before. Social Media is the fusion of technology and social behavior”
  • Toby Beresford: “… editorless content prioritised by popularity”
  • Jeff Jarvis: “Play is social. Media is play. Social media is fun”
  • Susan B. Barnes (Understanding Social Media from the Media Ecology Perspective): “… interpersonal media. It supports the sharing of personal exchanges in new and unique ways”
  • Bonnye E. Stuart, Marilyn S. Sarow, Laurence Stuart (Integrated Business Communication in a Global Marketplace): “… a term applied to digital media that enable customers to control content”
  • Wayne Kurtzman: “Social media is the use media, including and not limited to text, video, audio and documents within communities where users can opt to consume or generate content.”
  • Marc Smith: “Collective Goods produced through Computer-Mediated Collective Action”.

I’m planning to put together a systematic overview à la Kroeber/Kluckhohn. But to do this, I’m still looking for further ways to define social media. So, whose definition is still missing?

Obama and the global web

Global web? Sounds like redundancy, but in fact many phenomena on the world wide web aren’t so global at all. Although non-western languages are of rising importance online, English still is the most spoken language on the web. And so it should be clear that Barack Obama’s election is mostly an American or perhaps Western affair.

But this is not the case. If you take a look at Google search for “Obama” in the last 30 days, the USA only rank fourth on the list of countries searching for the president-elect:

  1. Ethiopia
  2. Kenya
  3. Cameroon
  4. Uganda
  5. United States
  6. Senegal
  7. Benin
  8. Ghana
  9. Nigeria
  10. Cuba

Is this a language issue – Obama perhaps being a very common name or word in those countries – or is this really a sign for the emergence of a more globalized internet? Is it about Obama or obama? There is some evidence for the second explanation: 1) the search volume clearly peaks on 5 November, the day after the election, 2) the three top related searches are for “barack obama”, “obama mccain” and “mccain”, 3) none of the mentioned countries were in the top list in 2005.

So, are we really witnessing the beginning of a truly global internet or a worldwide discursive space? Manuel Castells spoke of the emergence of global networks in production, communication, consummation and power, but also of locales or regions that are virtually cut off from this network. Maybe Africa is not so unlinked anymore?